Spatial Patterns of Urban Green-Blue Spaces and Residents’ Well-Being: The Mediating Effect of Neighborhood Social Cohesion

: Urban green-blue spaces (UGBS) can beneﬁt residents’ well-being through multiple pathways. Previous studies have conﬁrmed that the quantity and composition of UGBS can promote neighborhood social cohesion, which subsequently contributes to residents’ physical and mental health. However, there has been little attention paid to the spatial patterns of UGBS in such relationships. This study adopted landscape pattern indexes to characterize the spatial patterns of UGBS and explored the mediation effect of neighborhood social cohesion between the spatial patterns of UGBS and residents’ well-being, measured by self-rated health (SRH) and happiness. Partial Least Squares Structural Equation Model (PLS-SEM) was used for analyses with data obtained from the 2018 Shandong Provincial Social Survey Questionnaire (SGSS), which included 773 selected residents in urban areas. The results indicated that (1) there was a mediation effect of neighborhood social cohesion between the spatial patterns of UGBS and residents’ SRH and happiness; (2) the aggregation and diversity of UGBS had greater impacts on enhancing neighborhood social cohesion than the size, complexity, and fragmentation; (3) the aggregation and diversity of UGBS had indirect effects on improving happiness and SRH, and the aggregation of UGBS had a direct positive effect on SRH. By focusing on the spatial patterns of UGBS and neighborhood social cohesion, this study extends current debates on the pathways among UGBS, social cohesion, and public health. Urban planning strategies were proposed to increase the beneﬁts of UGBS in urban areas.


Introduction
Urban green space (UGS) is considered to be soft landscape elements, including trees, grass, and shrubs [1]. It mainly exists in parks, gardens, playgrounds, greenways, and woodlands [2]. Urban blue space (UBS) refers to lakes, springs, coastlines, canals, and ditches [3,4]. As important components of the urban system, UGS and UBS are beneficial to residents' well-being, which has often been discussed separately in previous studies [5][6][7]. Serving as outdoor spaces, urban green space and blue space are often considered an integrated entity by both residents and urban planners [8]. Individual elements in UGBS often work together to attract potential visitation of residents, and urban planners often arrange urban green space and blue space within a certain area simultaneously. Therefore, the distribution of urban green space and blue space is not independent; instead, they could affect residents' usage and subsequent health benefits as a whole [9]. To comprehensively capture the combined effects for urban green space and urban blue space on residents' wellbeing, recent studies started to consider them as a unified entity, known as urban green-blue space (UGBS) [10,11]. UGBS can function as an integrated buffer to mitigate the negative derived from 10-m resolution global land cover data in 2017 [49]. In order to describe the spatial patterns, size, and diversity of UGBS, landscape metrics (LMs) were applied. Partial least squares structural equation modeling (PLS-SEM) was conducted for estimation. As an exploratory study, the following questions were tested: (1) How can spatial patterns of UGBS affect residents' SRH and happiness? (2) Can neighborhood social cohesion play a role as a mediator between UGBS and well-being? (3) Which spatial pattern of UGBS is more directly linked to well-being? (4) Which spatial pattern of UGBS is more related to well-being when the mediating effects of neighborhood social cohesion are considered?

Data
The Shandong General Social Survey (SGSS), conducted by Shandong University from 2017 to 2019, was recently released for researchers. The SGSS surveyed residents who were over 18 years old and had lived in Shandong Province for more than half a year. This study utilized the 2018 wave of SGSS. A multi-stage sampling method was employed to select a total of 201 communities from 17 prefecture-level cities in Shandong Province, and the socioeconomic attributes of 3833 respondents, along with their community addresses, were recorded.
After removing samples located in rural areas and those with missing information, a total of 773 samples from 85 urban communities were selected for this study. We obtained the geographical coordinates of these urban communities in Shandong province using the mapping service of Gaode Map and matched these coordinates with the names of the communities in the 2018 SGSS dataset. We generated a 3 km circular buffer around each address of the 85 communities as the analytical unit and then calculated LMs to represent the spatial patterns of UGBS within these analytical units. The choice of a 3 km threshold is based on it being approximately a 15 min cycling distance, which can be considered the maximum range for residents to enjoy UGBS [50]. Additionally, 3 km is a comparable distance used in previous studies [34,51,52]. The addresses of sample communities, along with their buffers, are mapped in Figure 1.
as indicators to measure residents' well-being. Each respondent was asked a set of questions to evaluate the level of neighborhood social cohesion. Demographic and socio-economic characteristics were also collected in the survey. The analysis was conducted using a 3 km radius circle buffer around each location of the 85 communities. The types of UGBS were derived from 10-m resolution global land cover data in 2017 [49]. In order to describe the spatial patterns, size, and diversity of UGBS, landscape metrics (LMs) were applied. Partial least squares structural equation modeling (PLS-SEM) was conducted for estimation. As an exploratory study, the following questions were tested: (1) How can spatial patterns of UGBS affect residents' SRH and happiness? (2) Can neighborhood social cohesion play a role as a mediator between UGBS and well-being? (3) Which spatial pattern of UGBS is more directly linked to well-being? (4) Which spatial pattern of UGBS is more related to well-being when the mediating effects of neighborhood social cohesion are considered?

Data
The Shandong General Social Survey (SGSS), conducted by Shandong University from 2017 to 2019, was recently released for researchers. The SGSS surveyed residents who were over 18 years old and had lived in Shandong Province for more than half a year. This study utilized the 2018 wave of SGSS. A multi-stage sampling method was employed to select a total of 201 communities from 17 prefecture-level cities in Shandong Province, and the socioeconomic attributes of 3833 respondents, along with their community addresses, were recorded.
After removing samples located in rural areas and those with missing information, a total of 773 samples from 85 urban communities were selected for this study. We obtained the geographical coordinates of these urban communities in Shandong province using the mapping service of Gaode Map and matched these coordinates with the names of the communities in the 2018 SGSS dataset. We generated a 3 km circular buffer around each address of the 85 communities as the analytical unit and then calculated LMs to represent the spatial patterns of UGBS within these analytical units. The choice of a 3 km threshold is based on it being approximately a 15 min cycling distance, which can be considered the maximum range for residents to enjoy UGBS [50]. Additionally, 3 km is a comparable distance used in previous studies [34,51,52]. The addresses of sample communities, along with their buffers, are mapped in Figure 1.

Measuring Residents' Well-Being
To measure the well-being of urban residents, this study applied two indicators: SRH and happiness. SRH was assessed based on residents' responses to the question: "What is your current condition of health?" Respondents were provided with five options to select from: (1) extremely bad, (2) bad; (3) fair, (4) well, and (5) very well. Happiness was also

Measuring UGBS
Urban green and blue spaces were considered as a whole in this study and were extracted from a dataset of 10-m resolution land cover in 2017 [49]. The dataset included ten types of land cover, out of which forests, grassland, shrubland, freshwater, and wetland were selected as the components of UGBS.
LMs are commonly used to describe the size, spatial patterns, and diversity of UGBS [34]. In this study, we selected LMs from five dimensions of UGBS, as follows: (1) The size was measured using the largest patch index (LPI) and the percentage of landscape (PLAND). (2) The aggregation was reflected by the aggregation index (AI). (3) The degree of fragmentation was represented by the patch density (PD). (4) Complexity was assessed using the area-weighted mean fractal dimension index (FRAC_AM) and the areaweighted average shape index (SHAPE_AM). (5) Diversity was measured using Shannon's diversity index (SHDI). As mentioned earlier, the size and diversity dimensions correspond to the quantity and composition of UGBS, respectively, which have been widely confirmed to promote social cohesion and related health benefits for residents [30,31]. The other three dimensions encompass the key aspects of spatial patterns, including the clustering, distribution, and shape characteristics of UGBS; these aspects have high potential to affect residents social cohesion through promoting usability and social interactions [32][33][34][35][36][37]. All LMs were calculated using Fragstat 4.2 for each analytical unit [53]. The Supplementary Material provides detailed equations and descriptions for the LMs used in this study.

Mediating Variable: The Neighborhood Social Cohesion
Neighborhood social cohesion refers to the quality of life in the social environment formed by interpersonal communication or collective effort [54]. Indicators such as the sense of trust, belonging, acceptance, and relationships are commonly used to measure the level of neighborhood social cohesion [55][56][57]. In this study, we selected six indicators from the SGSS (2018) that represent six dimensions of neighborhood social cohesion. Respondents were asked if they agreed with the following statements: (1) "I can feel a sense of belonging to our community". (2) "I'm proud to tell people where I live". (3) "There is a harmonious relationship between different groups in our community". (4) "The residents of our community trust each other". (5) "In case of emergency, I can ask the residents in our community for help". (6) "The residents of our community love and help each other". The degree of agreement of the respondents was coded using a 5-item Likert scale. A value of 1 indicates "totally disagree", while a value of 5 indicates "total agree".
Age, marital status, having children or not, and personal annual income were considered covariates in this study. Age and personal annual income are continuous variables, while marital status was divided into two categories: married or other. Table 1 shows the descriptive statistics for all observed variables in this research.

Analytical Methods
Partial least squares structural equation modeling (PLS-SEM) was adopted in this study to explore the mediation effects, as it offers several benefits that are suitable for our case. Firstly, unlike other techniques such as multiple regression, PLS-SEM allows for the creation of complex models with numerous indicators to investigate mediation effects [54]. Secondly, PLS-SEM does not require assumptions about the distribution of the data and performs well with small sample sizes or missing data [54]. Thirdly, PLS-SEM utilizes efficient algorithms in model estimation, resulting in robust estimation results [54]. Lastly, the purpose of this study is exploratory; it investigates whether spatial patterns of UGBS may play various roles in enhancing neighborhood social cohesion and, consequently, residents' SRH and happiness.
PLS-SEM is made up of two elements: an outer model and an inner model. The outer model identifies the connections between latent variables, which are hardly measured directly, and their observed indicators and is therefore referred to as the "measurement model" [54]. The inner model, also known as the "structural model", further specifies the connections among latent variables [54]. In this study, SRH and happiness served as the two endogenous variables, while exogenous variables included the size, spatial patterns, and diversity of UGBS, neighborhood social cohesion, personal attributes, and economic level. To analyze the inner model, path coefficients were calculated and standardized to compare the relationships among latent variables, whose significance was tested using the bootstrapping method. To assess the outer model, three indicators, namely Cronbach's α, composite reliability (CR), and average variance extracted (AVE), were computed. Furthermore, standardized root mean square residual (SRMR) and normed fit index (NFI) were applied to evaluate the fitting ability of the whole model. The model validation results are discussed in detail in Section 4.2. SmartPLS3 was used for the analyses, and satellite-based data processing was completed on the ArcGIS platform.

PLS-SEM Validation
For the outer model, Cronbach's α, CR, and AVE were computed to test the validity and reliability of all constructs ( Table 2). All Cronbach's α and CR values exceeded the threshold of 0.6, indicating sufficient internal consistency and reliability in this study [58]. Additionally, the convergent validity of this model was acceptable, as all AVE values were above 0.5, which meant that the constructs explained over fifty percent of the variance in their indicators [58]. In terms of discriminant validity, each construct's AVE was higher than the squared correlation with other constructs. According to the Fornell-Larcker criterion, this suggested that all constructs were distinct from each other [58]. For the inner model, R 2 values were calculated to assess the predictive power of this model. The results pointed out that the R 2 values of SRH and happiness were 0.082 and 0.073, respectively. These values were consistent with similar studies on SRH and happiness, which typically ranged around 0.1 [59,60]. The variance inflation factor (VIF) values for the structural model were all below 5, indicating the absence of collinearity issues [58].
For the whole model, the SRMR and NFI were calculated to evaluate the overall fitting ability. The SRMR values for both models were below 0.1 (0.062 and 0.059), indicating a small difference between the observed and expected correlation matrices and thus indicating satisfactory model performance [61]. The NFI values for both models were relatively high (0.737 and 0.769), indicating acceptable model fitting ability [62]. Overall, the models employed in this study were deemed reasonable.  their indicators [58]. In the inner model, the connections between exogenous and endogenous variables are depicted by arrows and their corresponding values. Certain aspects of UGBS characteristics were significantly correlated with neighborhood social cohesion. The aggregation and diversity of UGBS had marginal positive effects on enhancing neighborhood social cohesion (β = 0.206, p < 0.01; β = 0.091, p < 0.05), while there were no significant relationships between the size, fragmentation, and complexity of UGBS and neighborhood social cohesion. Furthermore, levels of neighborhood social cohesion were positively associated with levels of residents' SRH (β = 0.123, p < 0.01). However, the size, spatial patterns, and diversity of UGBS did not have significant direct correlations with SRH. Additionally, personal attributes had negative effects on SRH (β = −0.258, p < 0.01), while economic level did not. Figure 2 displays the loading values of all indicators and path coefficients among latent variables, along with their corresponding p-values. In the outer model, all indicator loading values exceeded the recommended criterion of 0.7 and were statistically significant (p < 0.01), indicating that the constructs could be reliably and successfully represented by their indicators [58]. In the inner model, the connections between exogenous and endogenous variables are depicted by arrows and their corresponding values. Certain aspects of UGBS characteristics were significantly correlated with neighborhood social cohesion. The aggregation and diversity of UGBS had marginal positive effects on enhancing neighborhood social cohesion (β = 0.206, p < 0.01; β = 0.091, p < 0.05), while there were no significant relationships between the size, fragmentation, and complexity of UGBS and neighborhood social cohesion. Furthermore, levels of neighborhood social cohesion were positively associated with levels of residents' SRH (β = 0.123, p < 0.01). However, the size, spatial patterns, and diversity of UGBS did not have significant direct correlations with SRH. Additionally, personal attributes had negative effects on SRH (β = −0.258, p < 0.01), while economic level did not. When happiness entered as endogenous variable, the path model results are shown in Figure 3. In the outer model, the loading values of all indicators still exceeded the threshold of 0.7 with high significance. In the inner model, similar to the previous model, the aggregation and diversity of UGBS were positively correlated with neighborhood social cohesion (β = 0.200, p < 0.01; β = 0.090, p < 0.05), and neighborhood social cohesiveness and happiness were also positively correlated (β = 0.210, p < 0.01). However, compared with the previous model, the aggregation of UGBS had a direct positive impact on happiness (β = 0.115, p < 0.05), and unexpectedly, there was a negative association between the When happiness entered as endogenous variable, the path model results are shown in Figure 3. In the outer model, the loading values of all indicators still exceeded the threshold of 0.7 with high significance. In the inner model, similar to the previous model, the aggregation and diversity of UGBS were positively correlated with neighborhood social cohesion (β = 0.200, p < 0.01; β = 0.090, p < 0.05), and neighborhood social cohesiveness and happiness were also positively correlated (β = 0.210, p < 0.01). However, compared with the previous model, the aggregation of UGBS had a direct positive impact on happiness (β = 0.115, p < 0.05), and unexpectedly, there was a negative association between the size of UGBS and neighborhood social cohesion (β = −0.097, p < 0.1). The correlation between personal attributes and happiness remained highly significant, but the coefficient turned out to be positive (β = 0.074, p < 0.05). size of UGBS and neighborhood social cohesion (β = −0.097, p < 0.1). The correlation between personal attributes and happiness remained highly significant, but the coefficient turned out to be positive (β = 0.074, p < 0.05).

Influences and Total Effects
Tables 3 and 4 present the total and indirect effects, along with their effect sizes. The results of the indirect effects identified the mediation pathway of neighborhood social cohesion, through which the aggregation and diversity of UGBS were able to improve residents' SRH and happiness. The aggregation of UGBS had a positive impact on SRH and happiness through increasing neighborhood social cohesion, with indirect effects of 0.025 (p < 0.05) and 0.042 (p < 0.01), respectively. Similarly, the diversity of UGBS had indirect impacts of 0.011 (p < 0.1) and 0.019 (p < 0.05) on SRH and happiness. Regarding total effects, the effects of the aggregation and diversity of UGBS on happiness were 0.158 (p < 0.01) and 0.083 (p < 0.1). However, no significant total effects of the aggregation and diversity of UGBS on SRH were identified.
To determine the intensity of influence between a specific construct and endogenous constructs, effect size f 2 values were computed [58]. Based on previous studies, the intensity of influence can be categorized into three classes: small, medium, and large effects, with corresponding thresholds of 0.02, 0.15, and 0.35 [63]. As the dependent variable was SRH, personal attributes had a small negative effect on SRH (f 2 = 0.070), while neighborhood social cohesion did not have a large enough effect size (f 2 = 0.015). Furthermore, the effect sizes of the aggregation and diversity of UGBS on neighborhood social cohesion were both below 0.02 (f 2 = 0.014; f 2 = 0.006). In terms of happiness, neighborhood social cohesion had a small positive effect (f 2 = 0.045), while personal attributes and the aggregation and diversity of UGBS did not (f 2 = 0.006; f 2 = 0.005; f 2 = 0.003). The effect sizes of the aggregation and diversity of UGBS on neighborhood social cohesion remained below 0.02 (f 2 = 0.014; f 2 = 0.006).

Influences and Total Effects
Tables 3 and 4 present the total and indirect effects, along with their effect sizes. The results of the indirect effects identified the mediation pathway of neighborhood social cohesion, through which the aggregation and diversity of UGBS were able to improve residents' SRH and happiness. The aggregation of UGBS had a positive impact on SRH and happiness through increasing neighborhood social cohesion, with indirect effects of 0.025 (p < 0.05) and 0.042 (p < 0.01), respectively. Similarly, the diversity of UGBS had indirect impacts of 0.011 (p < 0.1) and 0.019 (p < 0.05) on SRH and happiness. Regarding total effects, the effects of the aggregation and diversity of UGBS on happiness were 0.158 (p < 0.01) and 0.083 (p < 0.1). However, no significant total effects of the aggregation and diversity of UGBS on SRH were identified.
To determine the intensity of influence between a specific construct and endogenous constructs, effect size f 2 values were computed [58]. Based on previous studies, the intensity of influence can be categorized into three classes: small, medium, and large effects, with corresponding thresholds of 0.02, 0.15, and 0.35 [63]. As the dependent variable was SRH, personal attributes had a small negative effect on SRH (f 2 = 0.070), while neighborhood social cohesion did not have a large enough effect size (f 2 = 0.015). Furthermore, the effect sizes of the aggregation and diversity of UGBS on neighborhood social cohesion were both below 0.02 (f 2 = 0.014; f 2 = 0.006). In terms of happiness, neighborhood social cohesion had a small positive effect (f 2 = 0.045), while personal attributes and the aggregation and diversity of UGBS did not (f 2 = 0.006; f 2 = 0.005; f 2 = 0.003). The effect sizes of the aggregation and diversity of UGBS on neighborhood social cohesion remained below 0.02 (f 2 = 0.014; f 2 = 0.006).  Note: * p < 0.1; ** p < 0.05; *** p < 0.01.

Discussion
Based on a province-wide survey conducted in Shandong Province, this study aimed to explore the mediation effect of neighborhood social cohesion between the spatial patterns, size, and diversity of UGBS within urban communities and residents' well-being. The results revealed the following findings: (1) there was a positive relationship between neighborhood social cohesion and both residents' SRH and happiness; (2) the aggregation of UGBS was positively associated with residents' happiness directly; (3) neighborhood social cohesion played a mediating role between the aggregation and diversity of UGBS and residents' SRH and happiness; (4) no significant correlation was captured between the size, fragmentation, and complexity of UGBS and residents' SRH and happiness.

Effects of UGBS and Neighborhood Social Cohesion on Residents' Well-Being
The significantly positive relationships between neighborhood social cohesion and residents' SRH and happiness are consistent with precedent research [64][65][66]. Neighborhood social cohesion can promote residents' SRH and happiness for several reasons. First, a greater level of neighborhood social cohesion represents a better sense of belonging and connection among residents, which contributes to maintaining mental health, reducing the risk of mental disorders, and subsequently increasing happiness [64]. Second, if residents are living in a neighborhood with a high social cohesion level, they are less likely to engage in behaviors that are detrimental to their own health and the health of others, such as smoking and spitting [65]. Last, neighborhood social cohesion level is often positively associated with frequent physical activities, which are beneficial for residents' physical health [66].
The aggregation, one important dimension of the spatial patterns of UGBS, was found to have a significant direct relationship with residents' happiness. The positive connection between the aggregation of UGBS and residents' happiness can be attributed to the following factors. Highly aggregated UGBS tends to consist of densely diverse vegetation and possibly include areas of different types of water bodies, creating an appealing landscape. Empirical evidence has found that visually appealing scenes can improve mental health through cognitive recovery and stress reduction [67]. Plus, a mix of clustered vegetation and water bodies can promote the flow of surrounding air due to their different thermal conductivity. The aggregation level of green and blue patches is known to have a strong capacity to lower surface temperature, which is an important factor in creating a sense of relaxation and reducing mental tension for residents [68][69][70]. Therefore, aggregated UGBS can directly contribute to residents' happiness through the creation of attractive sights, the reduction in local temperatures, and the promotion of air quality. Surprisingly, the aggregation and diversity of UGBS failed to improve SRH directly in our case. This may be because happiness is more closely related to mental measures, while SRH encompasses physical measures [43,71]. Mental stress can be partially alleviated through direct means such as observation and relaxation, but improving physical health requires the utilization of UGBS, which involves indirect approaches such as participating in activities. On the other hand, highly clustered and diverse UGBS may have direct detrimental effects on the SRH of individuals who are sensitive to negative factors such as pollen and mosquito bites [72,73].
Counterintuitively, the size of UGBS showed no significant relationship with residents' well-being, which aligns with the conclusions drawn in previous studies [34,51]. Their results suggest that the size of UGBS may not be directly proportional to the frequency of residents using UGBS, and thus the size of UGBS may not significantly contribute to residents' well-being [34]. The presence of a street view downstairs or a small nearby garden may only account for a small percentage of land cover, but they can indeed provide visual relief and spiritual comfort [51]. Conversely, if there is only one large UGBS with no other UGBS in the surrounding area, residents live further away may incur high costs to access the UGBS, making it less attractive, and the UGBS may not fully fulfill its potential in improving residents' well-being. Additionally, the quality of UGBS is another factor that can explain the mixed results [74]. Residents might also be willing to stay and enjoy small UGBS with various sports facilities and desirable environmental quality [75].

Mediation Effects of Neighborhood Social Cohesion between the Spatial Patterns and Diversity of UGBS and Residents' Well-Being
In this study, neighborhood social cohesion was identified as a mediator between the aggregation and diversity of UGBS and residents' well-being. The indirect effects of the aggregation of UGBS on residents' SRH and happiness can be attributed to several factors. Aggregated UGBS provides ample space for residents to meet with friends and neighbors for leisure and social communication, as well as opportunities for groups to organize various activities, which can promote neighborhood social cohesion [54]. UGBS with high aggregation can also provide attractive outdoor spaces for mental recovery and stress reduction, making residents more willing to visit and potentially increasing social interactions with each other [67]. Additionally, as noted above, highly clustered UGBS are associated with visually appealing aesthetics [76,77] and a comfortable breeze [68][69][70], creating a relaxing environment where residents are more inclined to participate in group physical activities and improve their SRH. Similarly, the indirect effects of the diversity of UGBS on residents' well-being arise from increased utilization of UGBS. UGBS with high diversity supports a wide range of fauna and flora, creating a more attractive space where residents actively engage in social gatherings and physical exercise [78]. A case study from Sheffield (UK) showed that indicators related to neighborhood social cohesion, such as positive emotional bonds, performed better as plant species richness increased [79]. Furthermore, high diversity in UGBS, with a more balanced proportion of various spaces, ensures that residents can carry out diverse activities [79].
Regarding the spatial patterns of UGBS, unlike the aggregation of UGBS, no significant direct or indirect connection was found between the fragmentation and complexity of UGBS and residents' well-being. Empirical studies have yielded counterproductive results regarding the effects of UGBS fragmentation on health or neighborhood social cohesion. Fragmented UGBS often has a dispersed distribution of various spaces, which may be more accessible to some residents [34]. However, a highly fragmented UGBS is often poorly visible, making it less appealing for residents to rest and participate in related group activities [80]. In addition, fragmented UGBS is often positively associated with air pollutants and negatively affects residents' physical health [24]. Regarding the complexity of UGBS, few studies have drawn significant conclusions about its positive effects on health. One study suggested that UGBS with high complexity may not make the best use of its internal space and boundaries, and reduce residents' exercise space, resulting in limited benefits to residents' SRH and happiness [77]. Besides, UGBS with high complexity may hinder temperature reduction by making it less favorable for residents to engage in group activities in outdoor areas and may neutralize possible health-related benefits [19].

Planning Recommendations
Based on the findings of this study, we propose three planning recommendations regarding the spatial distribution of UGBS to enhance the well-being of urban residents. Firstly, it is recommended to provide several clustered, large-scale UGBS to enhance neighborhood social cohesion and significantly improve residents' SRH and happiness. Planners should not only focus on increasing the overall quantity of UGBS but also adjust the spatial distribution by increasing the proportion of large parks and improving the aggregation of green spaces. Secondly, enhancing the diversity of UGBS is crucial for improving residents' well-being. Lastly, caution should be exercised when implementing dispersed and fragmented layouts of UGBS within a community, such as dividing large parks into multiple pocket parks, as excessively dispersed green space patterns may have negative impacts on residents' well-being.

Conclusions
This study utilized PLS-SEM to explore the mediation effects of neighborhood social cohesion between spatial patterns of UGBS and residents' SRH and happiness. The analyses incorporated data from the Shandong Social Survey (SGSS, wave 2018) and high-resolution (10 m) land use satellite imagery. The results revealed that: (1) neighborhood social cohesion was identified as a mediator between the spatial patterns of UGBS and residents' well-being, but not all spatial patterns worked the same way; (2) the aggregation and diversity of UGBS had a greater impact on improving neighborhood social cohesion compared with other indicators of spatial pattern; (3) the aggregation and diversity of UGBS had indirect effects on improving happiness and SRH, and the aggregation of UGBS had a direct positive effect on SRH. By focusing on the heterogenous effects of spatial patterns, this study provides further insights into previous research that primarily investigated the interrelationship among quantity/types of UGBS, neighborhood social cohesion, and residents' well-being. The results suggest that urban planners should prioritize the adjustment of the spatial distribution of UGBS and recognize the role of neighborhood social cohesion in order to promote residents' well-being.